Analysis of multiple datasets will be necessary to cover the full set of criteria, and to assess the information content for some individual criteria. The relative importance of each dataset ABT-263 in vivo is likely to be established by expert opinion. Datasets will almost certainly be at different spatial scales, and vary in their robustness and coverage. Datasets mapped either at a global scale or amalgamated from regional-scale sources are likely to

be necessary to provide comprehensive coverage of an area. It is important to be aware that datasets with broad areal coverage may contain sub-areas of low underlying data density, and/or sub-areas in which data values have been predicted using information from similar or adjacent areas. A check of underlying data should prevent misinterpretations, and indicate where high data density would support more detailed analysis if the management scale was smaller than the candidate EBSA identified. Where data are missing for certain criteria or where there are gaps in geographical coverage, the dataset or the criterion can be removed from consideration, or alternative options used to fill in the gaps (e.g., extrapolate from neighbouring areas, use proxy variables as a substitute,

expert opinion). These options will need to be evaluated on a case by case basis. As well as gathering Flucloronide appropriate

selleck chemical datasets, it may be necessary to set thresholds that reflect the intentions of the criteria. Whether an area meets the EBSA criteria mostly depends upon it exhibiting a comparatively “higher” value of diversity, productivity, vulnerability etc. than other areas. Determining the thresholds for each criterion requires an examination of the properties of the data being used. For example, the distribution of the data values may be such that exceptional sites will naturally stand out from others on histogram plots, and particular clusters or modes of data can be used to set a threshold. Expert knowledge should be used to interpret and justify the ecological validity of such data values, and in some instances statistical techniques can be used to identify the precise threshold value. For example, if the data distribution corresponds to standard models such as a normal distribution, sites can be identified using cut-offs at common statistical boundaries like quartiles, 95 percentile, or one or two standard deviations from the mean (Ardron et al., 2009). Data for the deep sea are generally sparse, and so pragmatic decisions will need to be made when determining appropriate datasets and thresholds. Notwithstanding any limitations, it is important that the properties of the datasets are fully described, and that threshold values are documented.